WO2018131733A1

WO2018131733A1 - Method and apparatus for reducing noise of ct image

Info

Publication number: WO2018131733A1
Application number: PCT/KR2017/000437
Authority: WO
Inventors: 김종효
Original assignee: 서울대학교산학협력단
Priority date: 2017-01-13
Filing date: 2017-01-13
Publication date: 2018-07-19

Abstract

A method for reducing noise may comprise the steps of: generating a synthetic sinogram from an inputted original CT image; obtaining a noise component from the generated synthetic sinogram; generating a noise component CT image on the basis of the noise component; and reducing noise of the original CT image on the basis of the noise component CT image.

Description

Noise reduction method and device of CT image

The present invention relates to a method and apparatus for reducing noise of a CT image.

Computed tomography (CT) can be taken by entering a large, circular machine with an X-ray generator to obtain a cross-sectional view across the human body, and the structure is less overlapping than simple X-rays. It is clearly seen that it is widely used in the examination of most organs and diseases.

The quality (resolution or precision) of CT images is a very important factor in the accurate diagnosis of lesions, and the development of CT systems continues to improve the quality of CT images. Multi-channel detector technology and high-speed high-resolution image reconstruction are also part of this effort. However, most efforts to improve the quality of CT images can lead to high doses of radiation exposure and are of concern. In particular, in view of recent social awareness of radiation exposure, efforts to obtain high quality diagnostic images should involve efforts to minimize radiation dose.

As part of this effort, CT manufacturers are launching low-exposure, high-quality CT systems. However, low exposure, high quality CT system does not allow easy access due to the high price compared to existing products and the difficulty of processing existing products. As an alternative to each effort, each of the CT manufacturers has been able to acquire high-quality CT images with low exposure through hardware / software upgrades to their existing products. However, this also can not be a real solution given the significant upgrade cost, and as a solution for this, I tried to develop the technology. Background art of the present application is disclosed in Korea Patent Publication No. 10-2014-0130784.

The present invention is to solve the above-described problems of the prior art, characterized in that for outputting a high-quality noise-reduced CT image from the input of a low-quality (resolution or precision) low-exposure CT image, wherein the noise reduction It is an object of the present invention to provide a method and apparatus for reducing noise in a CT image, which can show a high quality (eg, resolution or precision) that is comparable to that of a high-exposure CT image.

In addition, the present application is to provide a method and apparatus for reducing noise of the CT image to generate a composite sinogram from the input low-exposure CT image, and obtain a noise component image for the generated synthesized sinogram.

In addition, the present application obtains a noise component CT image by applying a filtered back projection operation to the noise component image obtained from the synthesized sinogram, and uses the noise of the CT image to generate a noise-reduced CT image using the same. It is intended to provide an abatement method and apparatus.

However, the technical problem to be achieved by the present embodiment is not limited to the technical problems as described above, and other technical problems may exist.

As a technical means for achieving the above technical problem, the noise reduction method according to an embodiment of the present application to generate a synthesized sinogram from the input original CT image, and synthesizes the noise component from the generated synthesized sinogram Obtaining a sinogram, generating a noise component CT image based on the noise component synthesis sinogram, and reducing the noise of the original CT image based on the noise component CT image have.

According to an example of the present embodiment, the generating of the synthesized sinogram may include attenuation coefficient for each pixel of the original CT image, distance information between an x-ray tube focus, and a detector based on the medical image information of the original CT image. Determining distance information between the x-ray tube focus and the patient and between the synthesis based on the determined pixel-specific attenuation coefficient, distance information between the x-ray tube focus and the detector and distance information between the x-ray tube focus and the patient. Generating a nogram.

According to an example of this embodiment, the synthesized sinogram is a projection by rotation angle based on the determined per-pixel attenuation coefficient, distance information between x-ray tube focus and detector, and distance information between x-ray tube focus and patient. Can be generated by performing an operation.

According to an embodiment of the present disclosure, obtaining a noise component synthesis sinogram in the synthesis sinogram may include obtaining a first noise component synthesis sinogram through noise component extraction in the synthesis sinogram; Extracting structural components in the first noise component synthesis sinogram and generating a second noise component synthesis sinogram from the first noise component synthesis sinogram by suppressing the extracted structural components. Can be.

According to an example of this embodiment, obtaining the noise component synthesis sinogram comprises extracting the noise component using at least one of a plurality of schemes, wherein the plurality of schemes include: The first method of determining the filter kernel according to a predetermined rule in gram and extracting the noise component based on this kernel, the second method of extracting the noise component based on the two-dimensional Fourier transform, based on the two-dimensional wavelet transform And a fourth method of extracting a noise component and a fourth method of extracting a noise component based on eigen decomposition of a Hessian matrix.

According to an example of the present embodiment, generating the noise component CT image based on the noise component synthesis sinogram may include generating a noise component CT image by applying a filtered backprojection operation to the noise component synthesis sinogram. It may include a step.

According to an example of the present embodiment, generating the noise component CT image may include generating a first noise component CT image by applying a reverse projection operation filtered to the noise component synthesis sinogram, and generating the first noise component. Extracting a structural component from a CT image and generating a second noise component CT image from the first noise component CT image by suppressing the extracted structural component.

According to an embodiment of the present disclosure, reducing the noise of the original CT image may include reducing the noise of the original CT image based on the noise component CT image.

According to an embodiment of the present disclosure, the reducing of the noise of the original CT image may include extracting tissue information from the noise component CT image and reducing noise of the original CT image based on the extracted tissue information. It may include.

According to an example of this embodiment, the step of reducing the noise of the original CT image, the noise of the original CT image by adaptively subtracting the noise component CT image from the original CT image based on the extracted tissue information It may include reducing the.

According to an example of the present embodiment, extracting a structural component from the noise component sinogram and the noise component CT image extracts the structural direction and the signal coherence for each pixel of the noise component sinogram and the noise component CT image. It may include a step.

According to an example of this embodiment, the structure direction of each pixel is a vertical direction of the normalized gradient vector in each pixel, and the signal coherence is the absolute value of the gradient value of the normalized gradient vector and the normalized gradient vector. It may be determined based on the absolute value of the inclination value of the vertical direction vector.

According to an example of this embodiment, the pixel-by-pixel structure direction is the direction of the second eigenvector of the Hessian matrix in each pixel, and the signal coherence is the two intrinsic of the Hessian matrix in each pixel. It can be determined based on the absolute values of the value.

According to an example of this embodiment, the structural direction and the signal coherence are determined based on a ratio between the absolute value of the slope of each pixel and the absolute value of the first eigenvalue of the Hessian matrix at each pixel, wherein the ratio When the reference value is larger than the reference value, the structure direction is a vertical direction of the normalized gradient vector in each pixel, and the signal coherence is an absolute value of the gradient value of the normalized gradient vector and the slope of the vertical direction vector of the normalized gradient vector. The structure direction is the direction of the second eigenvector of the Hessian matrix in each pixel, and the signal coherency is determined based on the absolute value of the value. It can be determined based on the absolute values of the two eigenvalues of the Hessian matrix at.

According to an example of the present embodiment, the extracting of the structural components from the noise component synthesis sinogram and the noise component CT image based on the structural direction and the signal coherence may include a two-dimensional ratio reflecting the structural direction and the signal coherence. The method may include determining a kernel corresponding to an isotropic Gaussian function and convolving the anisotropic kernel to each pixel of the noise component synthesis sinogram and the noise component CT image.

According to an example of this embodiment, the magnitude of the long axis among the parameters of the two-dimensional anisotropic Gaussian function is a predetermined value, and the magnitude of the short axis among the parameters is the magnitude of the long axis and the signal coherence and the predetermined proportionality constant. Determined by the product, the rotation angle of the parameter may be the structural direction.

As a technical means for achieving the above technical problem, the noise reduction device according to an embodiment of the present application and the synthesized sinogram generating unit for generating a synthesized sinogram from the input original CT image, and the generated synthesized sino A noise component acquisition unit for obtaining a noise component synthesis sinogram from a gram, a noise component CT image generator for generating a noise component CT image based on the noise component synthesis sinogram, and the noise component CT image based on the noise component CT image It may include a noise reduction unit for reducing the noise of the original CT image.

According to an example of this embodiment, the noise component synthesis sinogram acquisition unit may obtain the noise component synthesis sinogram through noise component extraction from the synthesis sinogram.

According to an example of this embodiment, the noise component CT image generator may generate the noise component CT image by applying a filtered backprojection operation to the noise component synthesis sinogram.

According to an example of the present embodiment, the noise reduction unit may extract tissue information from the original CT image and reduce noise of the original CT image based on the extracted tissue information.

According to an example of this embodiment, the noise reduction unit may reduce the noise of the original CT image by adaptively subtracting the noise component CT image from the original CT image based on the extracted tissue information.

The above-mentioned means for solving the problems are merely exemplary and should not be construed as limiting the present application. In addition to the exemplary embodiments described above, there may be additional embodiments described in the drawings and detailed description of the invention.

According to the above-described problem solving means of the present invention, characterized in that for outputting a high quality noise-reduced CT image from the input of a low quality (resolution or precision) low-exposure CT image, wherein the noise-reduced CT image is It can show a high quality (eg, resolution or precision) comparable to that of a high exposure CT image.

In addition, the present application may generate a synthesized sinogram from the input low-exposure CT image, and obtain a noise component synthesized sinogram from the generated synthesized sinogram.

In addition, the present application can generate noise component CT images through filtered backprojection on noise component synthesis sinograms.

In addition, the present application can output a high quality noise reduced CT image by reducing noise based on the original CT image and the noise component CT image.

In addition, the effects obtainable herein are not limited to the effects mentioned above, and other effects not mentioned may be clearly understood by those skilled in the art from the following description. will be.

1 is an overall conceptual diagram of a noise reduction apparatus according to an embodiment of the present application.

2 is a view showing the configuration of a noise reduction device according to an embodiment of the present application.

3A to 3C are diagrams illustrating a method of extracting a structure direction and signal coherence for each pixel according to an exemplary embodiment of the present application.

4 shows an anisotropic Gaussian kernel.

5 is a flowchart illustrating a noise reduction method according to an exemplary embodiment of the present application.

6 is a diagram illustrating a process of obtaining a noise component synthesis sinogram according to an embodiment of the present application.

7 is a diagram illustrating a process of extracting structural components from a noise component CT image according to an embodiment of the present disclosure.

DETAILED DESCRIPTION Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings so that those skilled in the art may easily implement the present disclosure. As those skilled in the art would realize, the described embodiments may be modified in various different ways, all without departing from the spirit or scope of the present invention. In the drawings, parts irrelevant to the description are omitted for simplicity of explanation, and like reference numerals designate like parts throughout the specification.

Throughout this specification, when a portion is "connected" to another portion, this includes not only "directly connected" but also "electrically connected" with another element in between. do.

Throughout this specification, when a member is located “on” another member, this includes not only when one member is in contact with another member but also when another member exists between the two members.

Throughout this specification, when a part is said to "include" a certain component, it means that it can further include other components, without excluding the other components unless specifically stated otherwise.

As used throughout this specification, the terms "about", "substantially" and the like are used at, or in the sense of, numerical values when a manufacturing and material tolerance inherent in the stated meanings is indicated, Accurate or absolute figures are used to assist in the prevention of unfair use by unscrupulous infringers. As used throughout this specification, the term "step to" or "step of" does not mean "step for."

In the present specification, the term 'unit' includes a unit realized by hardware, a unit realized by software, and a unit realized by both. In addition, one unit may be realized using two or more pieces of hardware, or two or more units may be realized by one piece of hardware.

Some of the operations or functions described as being performed by the terminal, the apparatus, or the device may be performed instead in the server connected to the terminal, the apparatus, or the device. Similarly, some of the operations or functions described as being performed by the server may be performed by the terminal, apparatus or device connected to the server. Hereinafter, an embodiment of the present disclosure will be described in detail with reference to the accompanying drawings.

Each configuration of FIG. 1 may be connected via a network. In this case, the network refers to a connection structure capable of exchanging information between respective nodes such as a plurality of terminals and servers, and examples of such a network include a 3rd Generation Partnership Project (3GPP) network and a Long Term Evolution (LTE). Network, World Interoperability for Microwave Access (WIMAX) Network, Internet, Local Area Network (LAN), Wireless Local Area Network (WLAN), Wide Area Network (WAN), Personal Area Network (PAN), Bluetooth (Bluetooth) ) Networks, satellite broadcasting networks, analog broadcasting networks, DMB (Digital Multimedia Broadcasting) networks, and the like.

As shown in FIG. 1, the noise reduction apparatus 100 receives a low dose CT image from the CT system 50 and generates a composite sinogram through projection based on the received CT image. The noise reduction apparatus 100 extracts a noise component from the generated synthesized sinogram, and performs noise reduction using the extracted noise component. Therefore, the noise reduction apparatus 100 may output the noise reduced image.

The noise reduction device 100 outputs a high quality noise reduced CT image from the input of the low exposure CT image, wherein the noise reduced CT image is compared with that of the high exposure CT image. High quality (e.g., resolution or precision).

2 is a view showing the configuration of a noise reduction device according to an embodiment of the present application. Referring to FIG. 2, the noise reduction apparatus 100 includes a synthesis sinogram generator 110, a noise component acquirer 120, a noise component CT image generator 130, and a noise reducer 140. . However, since the noise reduction apparatus 100 of FIG. 1 is only an example of the present disclosure, according to various embodiments of the present disclosure, the noise reduction apparatus 100 may be configured differently from FIG. 1.

Hereinafter, each configuration of the noise reduction apparatus 100 will be described in detail with reference to FIG. 2.

The synthesized sinogram generator 110 may generate a synthesized sinogram from the input original CT image.

In addition, the synthesized sinogram generating unit 110 is based on the medical image information of the original CT image, the attenuation coefficient for each pixel of the original CT image, the distance information between the x-ray tube focus and the detector and the distance between the x-ray tube focus and the patient Information can be determined.

At this time, when the synthesized sinogram generation unit 110 obtains the tube voltage information corresponding to the imaging of the original CT image based on the medical image information of the original CT image, the synthesized sinogram generation unit 110 for each pixel based on the tube voltage information and the attenuation coefficient table for each human tissue. Attenuation coefficients may be determined, and distance information between the x-ray tube focus and the detector and distance information between the x-ray tube focus and the patient may be determined based on the medical image information of the original CT image.

In this regard, the synthesized sinogram generator 110 may generate a synthesized sinogram based on the determined attenuation coefficient for each pixel, distance information between the x-ray tube focus and the detector, and distance information between the x-ray tube focus and the patient. have. In this case, the synthesized sinogram may be generated by performing projection operation for each rotation angle based on the determined pixel-specific attenuation coefficient, distance information between the x-ray tube focus and the detector, and distance information between the x-ray tube focus and the patient. .

The noise component acquirer 120 may obtain a noise component synthesized sinogram by extracting a noise component from the synthesized sinogram generated by the synthesized sinogram generator 110.

Specifically, the noise component obtaining unit 120 determines the noise size of each pixel of the virtual sinogram, extracts the structure direction and the signal coherence of each pixel of the virtual sinogram, and extracts the structure direction, the signal coherence and the noise size. Anisotropic bilateral filtering may be performed on the virtual sinogram based on the method, and the noise reduction filtered virtual sinogram may be generated by subtracting the anisotropic bilaterally filtered virtual sinogram from the virtual sinogram.

The noise component acquirer 120 may determine a filter kernel according to a rule specified in advance in the synthesized sinogram generated by the synthesized sinogram generator 110, and extract the noise component based on this. In addition, the noise component acquirer 120 may extract a noise component based on a two-dimensional Fourier transform, and may extract a noise component based on a two-dimensional wavelet transform. The noise component acquirer 120 may extract the noise component based on the eigen component decomposition of the Hessian matrix.

In detail, the noise component acquirer 120 uses the feature that the local change of the noise component is larger than the local change of the structural component, so that the filter kernel is set according to a predetermined rule to facilitate separation of the noise component and the structural component. The kernel can then filter the synthesized sinogram to extract noise components from the synthesized sinogram.

Specifically, the noise component obtaining unit 120 uses a feature in which the noise component is located in the high frequency band in comparison to the structural component in the two-dimensional Fourier transform region of the synthesized sinogram, thereby converting the synthesized sinogram to the two-dimensional Fourier transform and The noise component may be extracted from the synthesized sinogram by multiplying the band by a predetermined weight and then inverting the two-dimensional Fourier transform.

Specifically, the noise component acquisition unit 120 uses a feature that the noise component is located in the high frequency band compared to the structural component in the two-dimensional wavelet transform region of the synthesized sinogram, and converts the synthesized sinogram to the two-dimensional wavelet beforehand. After multiplying the weights by, the noise component can be extracted from the synthesized sinogram by inverse transforming the 2D wavelet.

Specifically, the Hessian matrix is a matrix of second-order partial derivatives in the vertical and horizontal directions in each pixel, and can be expressed as Equation (5), and the Hessian matrix H in the pixel (x, y). Since the first eigen component obtained when the eigen component is decomposed in [x, y] is a structural component, and the second eigen component represents a noise component, the noise component acquisition unit 120 is a In each pixel, the noise component may be extracted from the synthesized sinogram including the second eigen component of the Hessian matrix.

According to an embodiment of the present invention, the noise component acquisition unit 120 obtains the first noise component synthesis sinogram through noise component extraction from the synthesis sinogram generated by the synthesis sinogram generator 110. In addition, the structural component in the first noise component synthesis sinogram may be extracted. In addition, the noise component obtaining unit 120 may generate a second noise component synthesis sinogram from the first noise component synthesis sinogram by suppressing the extracted structural components.

The noise component CT image generator 130 may generate a noise component CT image based on the noise component synthesis sinogram obtained by the noise component acquirer 120.

In detail, the noise component CT image generator 130 may generate a noise component CT image by applying a filtered back projection operation to the noise component synthesis sinogram.

According to an embodiment of the present invention, the noise component CT image generator 130 may generate a first noise component CT image by applying a reverse projection operation filtered to the noise component synthesis sinogram. In addition, a second noise component CT image may be generated from the first noise component CT image by extracting a structural component from the first noise component CT image and suppressing the extracted structural component.

Hereinafter, a process of extracting structural components from the noise component sinogram and the noise component CT image by the noise component acquirer 120 and the noise component CT image generator 130 will be described.

In order to extract the structural components, the noise component acquirer 120 and the noise component CT image generator 130 may extract the structure direction and the signal coherence for each pixel from the sinogram and the original CT image, respectively. In this case, the structural direction may indicate a driving direction of the structure, and the signal coherence may be an indicator indicating how clear the direction of the signal structure is.

According to an embodiment of the present invention, the structural direction may be the vertical direction of the normalized gradient vector in each pixel, and the signal coherence is the absolute value of the gradient value of the normalized gradient vector and the slope of the vertical direction vector of the normalized gradient vector It can be determined based on the absolute value of the value.

A method of extracting the structure direction and the signal coherence for each pixel will be described with reference to FIGS. 3A to 3C.

Specifically, the driving direction vector Dg [x, y] of the structure having the inclined plane is obtained by obtaining the inclination vector G [x, y] as in Equation (1) at the given pixel position [x, y]. After normalizing together, the vertical direction can be obtained as Equation (3). At this time, the coherence Cg [x, y] of the signal structure can be obtained from the signal inclination value μ1 according to the normalized inclination vector and the signal inclination value μ2 in the vertical direction thereof. The preferred embodiment is shown in Equation (4). . (See step S30 to step S33 of FIG. 3A)

[Equation 1]

[Equation 2]

[Equation 3]

[Equation 4]

According to another embodiment of the invention, the structural direction is one of the directions of the eigenvectors of the Hessian matrix in each pixel, and the signal coherence is based on the absolute values of the two eigenvalues of the Hessian matrix in each pixel. It may be determined by.

Specifically, the structure direction may determine the second eigenvector V2 as the structure direction Dh [x, y] from the Hessian matrix H [x, y] as shown in Equation (5), and the signal coherence Ch [x, y]. ] Is determined as a result of dividing the difference between the absolute value of the first eigenvector and the absolute value of the second eigenvector by the sum of the absolute value of the first eigenvector and the absolute value of the second eigenvector. Can be. (See step S10 to step S14 of FIG. 3B)

[Equation 5]

[Equation 6]

According to another embodiment of the present invention, as shown in FIG. 3C, the structural direction and signal coherence are based on the ratio between the absolute value of the slope of each pixel and the absolute value of the first eigenvalue of the Hessian matrix at each pixel. It may be determined (S313).

At this time, when the ratio is larger than the reference value (S314), the structure direction is determined in the vertical direction of the normalized gradient vector in each pixel, and the signal coherence of the absolute value of the gradient value of the normalized gradient vector and the normalized gradient vector is determined. The determination can be made based on the absolute value of the inclination value of the vertical direction vector (see steps S30 to S33 in FIG. 3C).

Further, when the ratio is smaller than the reference value (S314), the structure direction is determined in the direction of the second eigenvector of the Hessian matrix in each pixel, and the signal coherence is two intrinsic of the Hessian matrix in each pixel. The determination may be made based on the absolute values of the value (see steps S11 to S14 of FIG. 3C).

In other words, as shown in Equation (7), when the ratio between the absolute value of the inclination in each pixel and the absolute value of the first eigenvalue of the Hessian matrix in each pixel is larger than the reference value T, the structure direction is determined in each pixel. If the ratio between the absolute value of the slope at each pixel and the absolute value of the first eigenvalue of the Hessian matrix at each pixel is less than or equal to the reference value T, the structure direction is determined for each pixel. It can be determined by the direction of the second eigenvector of the Hessian matrix at.

[Equation 7]

Also, as shown in Equation (8), when the ratio between the absolute value of the inclination in each pixel and the absolute value of the first eigenvalue of the Hessian matrix in each pixel is larger than the reference value T, the gradient of the signal is normalized. Is determined based on the absolute value of the absolute value of the inclination value of and the absolute value of the inclination value of the normalized vertical vector of the inclination vector, and the ratio between the absolute value of the inclination in each pixel and the absolute value of the first eigenvalue of the Hessian matrix in each pixel If less than or equal to the reference value T, it may be determined based on the absolute values of two eigenvalues of the Hessian matrix in each pixel.

[Equation 8]

In addition, according to an embodiment of the present invention, the noise component acquisition unit 120 has a structure direction and signal coherence according to equations (3) to (4) for an image having no or no linear structure according to the type of image. For the image with many linear structures, the structure direction and signal coherence are obtained according to equations (5) to (6), and for the intermediate image, the pixel is obtained according to equation (7) and equation (8). Alternatively, the structural direction and signal coherence can be determined selectively.

The noise component acquirer 120 and the noise component CT image generator 130 may perform anisotropic filtering on the noise component synthesized sinogram and the noise component CT image based on the structural direction and the signal coherence, respectively. Specifically, anisotropic kernels corresponding to the two-dimensional anisotropic Gaussian function reflecting the structure direction and signal coherence for each pixel may be determined, and filtering may be performed to reflect the anisotropic kernel. At this time, the magnitude of the long axis among the parameters of the two-dimensional anisotropic Gaussian function reflecting the structural direction and the signal coherence is a predetermined value, and the magnitude of the short axis among the parameters is the product of the magnitude of the long axis, the signal coherence and the predetermined proportionality constant. The rotation angle of the parameter may be a structural direction. In addition, the result of the anisotropic filtering may be a structural component of the noise component synthesis sinogram and the noise component CT image.

4 shows an anisotropic Gaussian kernel.

Referring to Equation (9), an anisotropic two-dimensional Gaussian function having long and short axis lengths of σx and σy, respectively, and an angle θ may be expressed as anisotropic by varying the length of the long and short axes. In addition, the anisotropic two-dimensional Gaussian function can express the degree of anisotropy by varying the ratio of the long axis and the short axis length, and may be suitable for generating an angled kernel kernel.

Given that the size of the anisotropic filter kernel is N, the long axis length of the two-dimensional anisotropic Gaussian function is σx = N, the short axis length is σx = (1-C (x, y)) N, and the angle

As described above, it is possible to generate a kernel in the form of an anisotropic two-dimensional Gaussian function using the direction and cohesion of the signal structure.

[Equation 9]

At this time,

As described above, the noise component obtaining unit 120 and the noise component CT image generating unit 130 perform anisotropic filtering based on the structural direction and the signal coherence of each pixel, respectively, to synthesize the noise component synthesized sinogram and the noise component, respectively. Structural components can be extracted from CT images.

In this case, a kernel may be generated by calculation for each pixel, and kernels corresponding to various signal direction and coherence of various signals are generated in advance in order to reduce the amount of calculation, and the signal structure direction and coherence obtained for each signal may be referred to as necessary. It can also be used by invoking the kernel.

The noise reduction unit 140 may reduce the noise of the original CT image based on the noise component CT image generated by the noise component CT image generator 130. In this case, the noise reduction unit 140 may reduce noise of the original CT image in various ways.

According to an example, the noise reduction unit 140 reduces noise of the original CT image by subtracting each pixel value of the noise component CT image corresponding to each pixel value of the original CT image from each pixel value of the original CT image. can do.

According to another example, the noise reduction unit 140 extracts tissue information (a range of previously known attenuation values for active ingredients, tissues, or organs) from the original CT image and based on the extracted tissue information, based on the extracted tissue information. Noise in the image can be reduced. At this time, the noise reduction unit 140 may reduce the noise of the original CT image by adaptively subtracting the noise component CT image from the original CT image based on the extracted tissue information. For example, the noise reduction unit 140 may reduce the degree of noise reduction in the region corresponding to the specific organization information.

According to another example, the noise reduction unit 140 may select a pixel whose pixel value is out of a predetermined range in the noise component CT image, and reduce the pixel value according to a predetermined rule, thereby avoiding damage to image quality. . For example, the noise reduction unit 140 selects only pixels having a pixel value equal to or greater than a predetermined multiple of the standard deviation calculated with respect to pixel values of all noise component pixels, or only pixels having pixel values of the upper 5% size. You can choose.

According to another example, the noise reduction unit 140 extracts the structural direction and signal coherence for each pixel of the original CT image, and determines a rule based on the structural direction, signal coherence and pixel values of the noise component CT image. Accordingly, the noise of the original CT image can be reduced.

In relation to the operation of the noise reduction unit 140, the process of extracting the structure direction and the signal coherence for each pixel of the original CT image may include the structural components of the noise component acquirer 120 and the noise component CT image generator 130. The same procedure as that used for extracting the ingredients is used, and thus the description thereof is omitted.

5 is a flowchart illustrating a noise reduction method according to an exemplary embodiment of the present application. The noise reduction method according to the embodiment shown in FIG. 5 includes steps processed in time series in the noise reduction device shown in FIG. 2. Therefore, although omitted below, the above description of the noise reduction apparatus shown in FIG. 1 may be applied to the noise reduction method according to the embodiment shown in FIG. 3.

In operation S100, the synthesized sinogram generator 110 of the noise reduction apparatus 100 may generate a synthesized sinogram from the input original CT image.

In addition, in step S100, the pixel-specific attenuation coefficient of the original CT image, the tube voltage of the x-ray tube, the distance information between the x-ray tube focus and the detector, and the distance information between the x-ray tube focus and the patient based on the medical image information of the original CT image. Determining may be further included.

The step S100 may further include generating a synthetic sinogram based on the determined attenuation coefficient for each pixel, distance information between the x-ray tube focus and the detector, and distance information between the x-ray tube focus and the patient.

In step S100, the synthesized sinogram may be generated by performing projection operation for each rotation angle based on the determined pixel-specific attenuation coefficient, distance information between the x-ray tube focus and the detector, and distance information between the x-ray tube focus and the patient. have.

Meanwhile, as described above, when the synthesized sinogram is generated in step S100, the noise component acquirer 20 of the noise reduction apparatus 100 may obtain the noise component synthesized sinogram from the generated synthesized sinogram ( S120).

In operation S100, a filter kernel is determined according to a predetermined rule in the synthesized sinogram, and the noise component is extracted based on the extracted filter kernel (S200). The noise component is extracted based on a two-dimensional Fourier transform (S210). And extracting the noise component based on the two-dimensional wavelet transform (S220) and extracting the noise component based on the eigen component decomposition of the Hessian matrix (S230).

Meanwhile, as described above, when the noise component synthesis sinogram is obtained from the synthesized sinogram in step S110, the image generator 30 of the noise reduction apparatus 100 generates a noise component CT image based on the noise component synthesis sinogram. It may be generated (S130).

In operation S130, the noise component CT image may be generated by applying a filtered backprojection operation to the noise component synthesis sinogram.

In addition, in step S130, generating a first noise component CT image by applying a reverse projection operation filtered to the noise component synthesis sinogram, extracting a structural component from the first noise component CT image, and extracting the extracted noise component. Generating a second noise component CT image from the first noise component CT image by suppressing a structural component.

When the noise component CT image is generated in step S130, the noise of the original CT image may be reduced based on the captured component CT image (S140).

Step S140 may include extracting tissue information from the original CT image and reducing noise of the original CT image based on the extracted tissue information and the noise component CT image.

In operation S140, the noise of the original CT image may be reduced by adaptively subtracting the noise component CT image from the original CT image based on the extracted tissue information.

In operation S140, the pixel value of the noise component CT image may be reduced according to a predetermined rule based on the distribution order of the pixel values of the noise component CT image pixels.

7 is a view showing a process of extracting a structural component according to an embodiment of the present application.

Extracting the structural component may extract the noise component using at least one of a plurality of methods. The plurality of methods perform a method of extracting the structural direction and signal coherence for each pixel of the original image (S300), a method of determining anisotropic kernel 310 based on the structural direction and signal coherence, and filtering the reflection of the anisotropic kernel. Method 320 may be included. Extracting these structural components may include extracting noise components using at least one of a plurality of methods. All methods can be used to extract noise components.

Each method described above (e.g., a method for reducing noise in a CT image) may also be implemented in the form of a recording medium containing instructions executable by a computer, such as a program module executed by the computer. Computer readable media can be any available media that can be accessed by a computer and includes both volatile and nonvolatile media, removable and non-removable media. In addition, computer readable media may include both computer storage media and communication media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Communication media typically includes computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave, or other transmission mechanism, and includes any information delivery media.

The above description of the present application is intended for illustration, and it will be understood by those skilled in the art that the present invention may be easily modified in other specific forms without changing the technical spirit or essential features of the present application. Therefore, it should be understood that the embodiments described above are exemplary in all respects and not limiting. For example, each component described as a single type may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a stacked form.

The scope of the present application is indicated by the following claims rather than the above description, and it should be construed that all changes or modifications derived from the meaning and scope of the claims and their equivalents are included in the scope of the present application.

Claims

In the method of reducing the noise of the CT image

Generating a synthesized sinogram from the input original CT image;

Obtaining a noise component synthesis sinogram from the generated synthesis sinogram;

Generating a noise component CT image based on the noise component synthesis sinogram; And

Reducing noise of the original CT image based on the noise component CT image;

Comprising a noise reduction method of the CT image.
The method of claim 1,

Generating the synthetic sinogram,

Determining a pixel-specific attenuation coefficient, distance information between an x-ray tube focus and a detector, and distance information between an x-ray tube focus and a patient of the original CT image based on the medical image information of the original CT image; And

Generating the synthesized sinogram based on the determined pixel-by-pixel attenuation coefficient, distance information between the x-ray tube focus and the detector, and distance information between the x-ray tube focus and the patient;

Comprising a noise reduction method of the CT image.
The method of claim 2,

The synthetic sinogram,

The noise of the CT image, characterized in that generated by performing the projection operation for each rotation angle based on the determined per-pixel attenuation coefficient, the distance information between the x-ray tube focus and the detector and the distance information between the x-ray tube focus and the patient Abatement method.
The method of claim 1,

Acquiring the noise component synthesis sinogram,

Obtaining a first noise component synthesis sinogram through noise component extraction from the synthesized sinogram;

Extracting structural components in the first noise component synthesis sinogram; And

Generating a second noise component synthesis sinogram from the first noise component synthesis sinogram by suppressing the extracted structural components;

Comprising a noise reduction method of the CT image.
The method of claim 1,

Acquiring the noise component synthesis sinogram,

Extracting a noise component using at least one of a plurality of schemes,

The plurality of methods,

A first method of determining a filter kernel according to a predetermined rule in the synthesized sinogram and extracting a noise component based on the kernel;

A second method of extracting noise components based on a two-dimensional Fourier transform,

A third method of extracting noise components based on two-dimensional wavelet transform, and

A fourth scheme of extracting noise components based on eigencomponent decomposition of a Hessian matrix;

Comprising a noise reduction method of the CT image.
The method of claim 1,

Generating the noise component CT image,

And generating a noise component CT image by applying a filtered backprojection operation to the noise component synthesis sinogram.
The method of claim 1,

Generating the noise component CT image,

Generating a first noise component CT image by applying a filtered backprojection operation to the noise component synthesis sinogram;

Extracting structural components from the first noise component CT image; And

Generating a second noise component CT image from the first noise component CT image by suppressing the extracted structural component;

Noise reduction method of the CT image comprising a.
The method of claim 1,

Reducing the noise of the original CT image,

Extracting tissue information from the original CT image; And

Reducing noise of the original CT image based on the extracted tissue information and the noise component CT image;

Noise reduction method of the CT image comprising a.
The method of claim 8,

Reducing the noise of the original CT image,

Reducing noise of the original CT image by adaptively subtracting the noise component CT image from the original CT image based on the extracted tissue information;

Noise reduction method of the CT image comprising a.
The method of claim 1,

Reducing the noise of the original CT image,

Reducing the pixel values of the noise component CT image according to a predetermined rule based on the distribution rank of the pixel values of the noise component CT image pixels

Noise reduction method of the CT image comprising a.
The method according to claim 4 or 7,

Extracting the structural component,

Extracting structure direction and signal coherence for each pixel of the original image;

Determining anisotropic kernels based on structure orientation and signal coherence; And

Performing filtering reflecting the anisotropic kernel;

Noise reduction method of the CT image comprising a.
The method of claim 11,

The structure direction is a vertical direction of a normalized gradient vector in each pixel,

And the signal coherency is determined based on an absolute value of an inclination value of the normalized gradient vector and an absolute value of an inclination value of the normalized gradient vector of the normalized gradient vector.
The method of claim 11,

The structural direction is the direction of the second eigenvector of the Hessian matrix in each pixel,

And said signal coherence is determined based on absolute values of two eigenvalues of a Hessian matrix in each pixel.
The method of claim 11,

Wherein the structural direction and the signal coherence are determined based on a ratio between the absolute value of the slope of each pixel and the absolute value of the first eigenvalue of the Hessian matrix at each pixel.

When the ratio is greater than a reference value, the structure direction is a vertical direction of the normalized gradient vector in each pixel, and the signal coherency is an absolute value of the slope value of the normalized gradient vector and the vertical direction vector of the normalized gradient vector. Is determined based on the absolute value of the slope of

If the ratio is less than a reference value, the structure direction is the direction of the second eigenvector of the Hessian matrix in each pixel, and the signal coherence is the absolute value of the two eigenvalues of the Hessian matrix in each pixel. Noise reduction method of a CT image, characterized in that determined based on the.
The method of claim 13,

Determining the anisotropic kernel based on the structural direction and the signal coherence is determined by the kernel based on the two-dimensional anisotropic Gaussian function, the size of the long axis of the parameters of the two-dimensional anisotropic Gaussian function is a predetermined value, The magnitude of the short axis of the parameter is determined by multiplying the magnitude of the long axis by the signal coherence and a predetermined proportional constant, wherein the rotation angle of the parameter is the structural direction, CT noise reduction method.
In the device that reduces the noise of the CT image

A synthesized sinogram generator for generating a synthesized sinogram from the input original CT image;

A noise component obtaining unit for obtaining a noise component synthesis sinogram from the generated synthesis sinogram;

A noise component CT image generation unit generating a noise component CT image based on the noise component synthesis sinogram; And

And a noise reduction unit for reducing noise of the original CT image based on the noise component CT image.
The method of claim 16,

The noise component obtaining unit,

And extracting a noise component from the synthesized sinogram to obtain the noise component synthesized sinogram.
The method of claim 16,

The noise component CT image generation unit,

And generating the noise component CT image by applying a filtered reverse projection operation to the noise component synthesis sinogram generated by the noise component obtaining unit.
The method of claim 16,

The noise component acquisition unit and the noise component CT image generation unit,

And extracting structural components from the noise component sinogram and noise component CT images, respectively, to suppress the extracted structural components.
The method of claim 16,

The noise reduction unit,

And extracting tissue information from the original CT image, and reducing noise of the original CT image based on the extracted tissue information.
The method of claim 20,

The noise reduction unit,

And reducing the noise of the original CT image by adaptively subtracting the noise component CT image from the original CT image based on the extracted tissue information.